New feature request: Stereo/photogrammetric imaging

[Bart_van_der_Wolf]

Hi,

One of the features that's useful for postprocessing is the depthmap that can be exported from the Windows version of HeliconFocus. It can be used for re-introducing a Depth of Field drop-off in sharpness which allows to produce a very natural looking image (although with a physically impossible Depth of Field). It can also be used for selective focussing to direct attention to various focus planes, or the creation of masks to isolate subjects from their surroundings.

This sparked my interest in expanding the usability of the Depth Map for measurements of distance/depth. As I've understood from a reaction by Stas to another post, the depthmap is encoded as graylevels (range 0..256, or 0 .. 65535) corresponding to the stack layer number, and normalized to cover the density range of the file format. As long as the input layers can be calibrated to known distances or a known displacement, this can work fine to make depth measurements. The combination with e.g. a Stackshot setup can achieve decent accuracy, although limited to mostly off-axis structures. There is a limited capability to look 'behind' thin object edges.

This made me wonder, and especially useful for more distant objects where DOF is large (and focus layers are few), wouldn't it be useful to allow HeliconFocus to use stereoscopic data to build a depth map? One could think of a simple approach by using the parallax generated by a known axial displacement of an image pair (or a focus stacked output pair), all the way to using the combination of Depth from stacking layer, and axial displacement. The simplest application would use parallel axes to shoot, after which a shift of the resulting images would set a zero parallax at a give distance, and all other relative distances would be calculated from the amount of parallax. A zero shift would equal a calibration at infinity. Such a setup is easy to calibrate, because only a simple shift between the axes need to be known, which is easier than having to also calibrate the axis convergence, and the actual or relative distance at 2 positions. Besides, a microscope also allows a simple X/Y translation and calibration between 2 positions, so this would be a universal approach for different applications.

I'm sure that that would also open up new usage opportunities for HeliconFocus, e.g. simple forensic research, or archeological/photogrammetric uses without the need for other dedicated software than HeliconFocus. That is in addition to the more creative uses of depth maps for controlling DOF in postproduction of e.g. product photography, the production of improved Masks to isolate subjects, or the production of improved 3D models compared to the basic ones that HeliconFocus already supports.

Cheers,
Bart

[Stas Yatsenko]

Thanks for your suggestions. What you are talking about is stereophotogrammetry (http://en.wikipedia.org/wiki/Photogrammetry) and is quite a different field from what Helicon Focus does now. You are right about use of depth map, it would be interesting to use it for selection and selective DOF, we will think how to implement this.